[gmueckl]

Cleaning up your own process between fork and exec is hard. Several programs resort to terrible hacks like force-closing everything except file IDs 0,1,2 in a loop. Or they look into their /proc directory to discover whichnfile IDs exist, which is only marginally better. But when your process is a house of cards built on third party libraries with their own minds, there are not a lot of other options.

[tobias3]

Use O_CLOEXEC everywhere (even third party libs). It's really annoying, but necessary. Means you need to use accept4(), dup3(), popen with an additional "e" (of course all of that needs to be feature tested, during compilation/runtime).

[gmueckl]

The catch is that you may not be able to control 3rd party libraries enough to be be able to do all that. Thus all these annoying hacks. To me, the complexity of using fork() and the race conditions around pid reuse are the worst design problems of POSIX systems.

[cryptonector]

Win32 has the opposite semantics, that O_CLOEXEC is the default semantics and the app has to request the opposite if it wants it, and this causes problems too. There should have been two flags and the application should have to specify one on every handle-/fd-creating system call. Hindsight is 20/20.